■
RESULTS
NaCl Dependence of DiI Aggregation. Although
lipophilic dyes are commercially available as powders or as
solutions in organic solvents such as ethanol, dimethylforma-
mide (DMF), or dimethyl sulfoxide (DMSO), they need to be
transferred to aqueous solutions for vesicle labeling not to
disrupt vesicles and the embedded proteins. This requirement
poses a challenge because the water solubility of lipophilic
molecules is generally low, as shown by their high lipophilicity
(calculated log P values for the octanol/water partition
coecient
15
are given in Figure 1). Therefore, we first directly
examined DiI molecules dissolved in aqueous solutions by
single-molecule TIRF microscopy
16
(Figure S1A). After
introducing DiI solutions into a glass flow cell, the fluorescent
particles floating by near the glass surface were illuminated.
We first imaged 2 μM DiI solution in a buer with ∼150
mM NaCl, a physiological and typical condition for common
buers including PBS. A small number of bright, slowly
moving particles were detected (Figure S1A), which are likely
large aggregates of DiI rather than single molecules. Since most
fluorescent dyes including DiI exhibits aggregation-caused
quenching of fluorescence intensity, the brightness of the
particles would actually underestimate the number of dye
molecules per particle. Indeed, these aggregates were
completely removed by filtering the solution through 0.2 μm
pores (Figure S1A,B), implying that they are mostly micron-
sized. These large aggregates are likely inecient at labeling
vesicular membranes and may cause an increase in vesicle size
after labeling.
14
We therefore attempted to improve the solubilization of DiI
by decreasing NaCl concentration. The aggregates gradually
dispersed as NaCl concentration was lowered to ∼20 mM, as
shown by the increase of relatively dim particles (Figure S1A−
C). Importantly, we checked that these changes to particles
occurred while the total amount of dye molecules and their
fluorescence remained constant: After solubilizing the dye
aggregates completely with detergent (0.1% Triton X-100), the
overall fluorescence intensity from the solution was measured
to be the same across the concentrations of NaCl we tested
(Figure S1D). In contrast, the fluorescence from the buer
with 155 mM NaCl almost completely disappeared after
micropore filtering (Figure S1D; ∼5 nM DiI left from the
original 2 μM solution), suggesting that most of the dye
molecules in this condition were trapped in the aggregates and
subsequently removed.
Improvement of Fluorescent Labeling by the Salt-
Change Method. The above results suggest that dispersion of
DiI in a buer with a low concentration of NaCl can potentiate
membrane partitioning of DiI and that the excess dye can be
removed by inducing its aggregation at a higher concentration
of NaCl. We therefore exploited this reversible aggregation of
DiI to improve the labeling of vesicles (Figure 2). To verify the
general applicability of labeling procedures, we employed cell-
derived vesicles (CDVs) as model EVs that are similar in size
to large exosomes and small microvesicles.
17
CDVs from
human natural killer cells (hereafter called NK-CDVs) were
prepared by using a published procedure,
18
and labeled them
with DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocya-
nine; DiIC
18
(3)), a lipophilic fluorescent tracer for labeling
lipid membranes.
For the side-by-side comparison of labeling eciency, two
labeling methods were applied to NK-CDVs: (a) “Direct
staining” performed with 150 mM NaCl, that is, in regular PBS
(Figure 2A), and (b) “Salt-change” approach in which the
labeling with and removal of DiI were performed separately in
buers with distinct ionic strength (Figure 2B). In the latter
method, after staining NK-CDVs in a low-salt buer ([NaCl] <
20 mM), we raised NaCl concentration to ∼150 mM to induce
aggregation of free dye molecules, then filtered the solution
using a regular syringe filter. This filtering step was applied also
to the direct staining procedure for a fair comparison of
labeling results and vesicle yield. The labeled CDVs were then
visualized using a TIRF microscope
16
(Figure 2C).
CDVs after direct staining displayed only a small number of
dim particles (Figure 2D; see also Video S1). Since the CDVs
were prepared at a fairly high concentration (∼10
10
particles/
mL), we expected much more particles to be present in the
field of view, and therefore, it was very unlikely that all CDVs
were successfully labeled by the direct staining method.
Although the observed level of labeling eciency might be
suitable for bulk assays that probe many vesicles at the same
time (e.g., cellular uptake of vesicles), the labeled CDVs were
neither suciently abundant nor suciently bright for
quantitative measurements at the single-vesicle level. Accord-
ing to our observations of NaCl concentration-dependent DiI
aggregation, we argued that the low labeling eciency would
stem from poor solubilization of lipophilic dyes in the staining
buer.
19
In stark contrast, the salt-change method increased
the number of bright fluorescent vesicles 85 ± 10 times
(Figure 2D,E; see also Video S1), and their average brightness
also increased 2.3 times compared to vesicles stained in PBS
with 150 mM NaCl (Figure 2F). The simultaneous increase in
number and brightness of fluorescent vesicles implies that the
overall DiI incorporation (estimated from the areas under the
curves in Figure 2F) was improved by a factor of 290.
To accurately measure the labeling density (i.e., number of
DiI molecules per vesicle), the labeled NK-CDVs were stably
captured on a surface and their fluorescence intensity was
measured (Figure S2A; see Supporting Information for the
method). We estimated the number of DiI molecules in each
vesicle from the ratio of the initial fluorescence to photo-
bleaching step size (Figure S2B−D). Each CDV typically
carried 1−3 molecules of DiI, and these numbers followed a
Poisson distribution as expected (Figure S2E). The results
imply that only a small fraction of the CDVs remained
unlabeled and, at the same time, that the mole fraction of DiI
in vesicle membranes was <10
−4
(less than 10 dye molecules vs
∼10
5
lipid molecules; see Supporting Information for the full
calculation). Therefore, the labeling density we achieved was
sucient for single-vesicle imaging, but unlikely to disrupt the
native properties of the membrane.
Applications to Other Vesicles and Dyes. To test
whether the salt-change labeling method can be applied to
other vesicles, we first prepared another sample of CDVs from
umbilical cord mesenchymal stem cells (UCMSC-CDVs) and
labeled them with DiI. Again, the salt-change method showed a
dramatic improvement in labeling eciency (Figure S3A,B),
consistent with the results for NK-CDVs. It is remarkable that
the proposed method was much more eective than adding
dimethyl sulfoxide (DMSO) to the staining buer (Figure
S3A,B), a common approach to improve the solubility of
lipophilic dyes. Also, if the syringe filters for dye removal were
not rinsed with buer (2 mL of PBS) before use, we noticed
that the vesicle yield decreased slightly (by 17%; Figure
Analytical Chemistry pubs.acs.org/ac Technical Note
https://doi.org/10.1021/acs.analchem.2c05166
Anal. Chem. XXXX, XXX, XXX−XXX
C